Grok 4.5 has received positive feedback for its performance in blockchain attack analysis, particularly when integrated with the Hermes Agent for harness engineering related to the SlowMist Team. The latest version has shown significant improvements over previous iterations, which had been disappointing to some users. The satisfaction hierarchy for various models and agents, as noted by a user, ranks GPT 5.5/5.6 and GLM 5.2 at the top, followed by Grok 4.5, and then DeepSeek V4 Pro/GLM 5.1. The user expressed that Claude is no longer essential due to its perceived limitations. The rapid iteration of models and agents suggests that practical experience will become the key differentiator in the field, as technical barriers continue to diminish.